Wednesday, June 22, 2011

Circles of Geeks

Circles of Geek As Internet Consulting editor for the Annals of Botany, I'm off to the Editorial Board meeting at the Royal Society (rather looking forward to that) today, where I will be expected to say something interesting, or at least not too embarrassing, about the Internet. As an icebreaker, each Editor has been asked to talk for two minutes about "an interesting paper". Not being a plant scientist, I've picked:

Yan K-K, Gerstein M, 2011 The Spread of Scientific Information: Insights from the Web Usage Statistics in PLoS Article-Level Metrics. PLoS ONE 6(5): e19917. doi:10.1371/journal.pone.0019917

There is some interesting analysis in this paper. Popular papers are downloaded more than less popular papers (this is the definition of popularity), but irrespective of popularity, downloads all decay with a consistent pattern:
  • Rapid decline within the first month of publication
  • Slower decline thereafter (a long tail)
Why does this dual pattern occur? In the memorable words of the authors:
nontrivial herding behavior analogous to what we referred to as fame
coupled with the Matthew effect - a stochastic model of information diffusion (Circles of Geeks). Clearly, the only thing worse than being talked about is not being talked about.



  1. Interesting. I was looking at some similar stuff recently too: Iribarren and Moro use the term 'affinity paths' to show that messages are more likely to be spread when the content has an affinity to the individual (not very surprising really - Szabo and Huberman have an interesting finding that for Digg you can predict popularity from the first two hours, whereas you need 10 days for YouTube, suggesting that different tools have different activation patterns, Digg being very immediate and YouTube taking longer to spread and catch on ( Perhaps germane to this article - Crane and Sornette reveal two patterns of social network spread - the early burst and the steady growth. So for example conversation about the tsunami rose very quickly and then decreased from this peak, whereas Harry Potter bubbled along increasing to a peak around news about the film, and slowly dissipated (
    I am also interested in how the online catalogue induces long tail behaviour - ie people start searching for more niche items. See Brynjolfsson, Hu & Simester (2007).
    maybe we should write a paper?

  2. I've sent you an email :-)